Threshold-based scene generation method, device and storage medium
By sampling the attribute values of the target object and filtering the constraints, the target threshold is determined to update the original scene, which solves the problem of low efficiency in virtual scene generation under manual editing and achieves efficient and accurate virtual scene generation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN DEEPROUTE AI CO LTD
- Filing Date
- 2022-07-29
- Publication Date
- 2026-07-03
AI Technical Summary
In existing technologies, when generating virtual scenes through manual editing, it is impossible to accurately determine the driving boundary threshold of the target object, resulting in low efficiency in virtual scene generation.
By acquiring the original scene file of the target object, performing attribute value sampling processing, generating a sampling set, filtering out the target sampling value group according to preset constraints, determining the target threshold, and then updating the original scene.
It achieves efficient generation of target scenes, improves the efficiency and accuracy of virtual scene generation, and generates richer scenes to meet diverse needs.
Smart Images
Figure CN115408822B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of autonomous driving simulation technology, and in particular to a threshold-based scene generation method, apparatus and storage medium. Background Technology
[0002] In the field of autonomous driving simulation technology, building virtual scenarios for road testing can ensure the correctness of autonomous driving algorithms and help identify problems in the testing process. Therefore, building virtual scenarios efficiently and accurately is a crucial step in conducting road tests.
[0003] Currently, multiple sets of attribute information are typically generated manually, and then combined with the target object to obtain the corresponding virtual scene. However, since the target object can only travel along a preset route, manual editing cannot determine the safe boundaries of the target object's travel. Therefore, how to accurately obtain the boundary thresholds of the target object during its travel and thus efficiently generate the virtual scene is the problem that this application aims to solve. Summary of the Invention
[0004] Therefore, it is necessary to provide a scene generation method, apparatus, and storage medium that can improve the construction efficiency of the target scene in order to address the above-mentioned technical problems.
[0005] Firstly, this application provides a threshold-based scene generation method. The method includes:
[0006] Obtain the original scene file containing the target object; the original scene file includes multiple attribute types of the target object, and the attribute value range corresponding to each attribute type;
[0007] Each attribute value range is sampled to obtain a sample set corresponding to each attribute type.
[0008] By combining the sampled values from different sample sets, multiple sampled value groups are obtained;
[0009] The target sample value group is selected from the multiple sample value groups according to the preset attribute value constraints;
[0010] Based on the target sample value group, the target threshold of the target object under the corresponding attribute type is determined, and the original scene is updated according to the target threshold to obtain the target scene.
[0011] In one embodiment, before obtaining the original scene file containing the target object, the method further includes: obtaining the original scene file; the original scene file includes multiple initial attribute types and attribute information corresponding to each initial attribute type; the attribute information includes at least one of attribute value range and attribute constant; performing field traversal on the attribute information corresponding to each initial attribute type; if the attribute information currently traversed includes an attribute constant, then determining that the initial attribute type is an attribute type that does not require threshold adjustment; if the attribute information currently traversed includes an attribute value range, then determining that the initial attribute type is an attribute type that requires threshold adjustment.
[0012] In one embodiment, the attribute value range includes a sampling interval, an original minimum value, and an original maximum value; the step of performing attribute value sampling processing on each of the attribute value ranges to obtain a sampling set corresponding to each attribute type includes: for each of the multiple attribute types, starting from the original minimum value, and sampling attribute values according to the sampling interval until the original maximum value is sampled, to obtain the sampling value corresponding to each sampling; and combining the multiple sampling values to obtain the sampling set corresponding to each attribute type.
[0013] In one embodiment, combining sampled values from different sample sets to obtain multiple sampled value groups includes: combining sampled value M from sample set i to sampled value n. (i，n) The sample value M in the sample set described by j is k. (j，k) By combining them, multiple sample value groups are obtained; i∈(1,2…N), j∈(1,2…N), N represents any one of the multiple sample sets; n, k, and N are all positive integers.
[0014] In one embodiment, the attribute value constraints are obtained through an autonomous driving evaluation system; the step of selecting a target sample value group from the plurality of sample value groups according to the preset attribute value constraints includes: for each of the plurality of sample value groups, determining whether the behavioral characteristics of the target object conform to preset traffic rules by using each sample value in the current sample value group corresponding to the target object; when the behavioral characteristics conform to the preset traffic rules, determining whether the behavioral characteristics of the target object will experience a sudden situation; the sudden situation includes at least one of collision and rear-end collision; when the sudden situation does not occur, the current sample value group is used as the target sample value group.
[0015] In one embodiment, determining the target threshold of the target object under the corresponding attribute type based on the target sample value group includes: clustering the sample values in the target sample value group according to the attribute type to obtain multiple target sample values corresponding to each attribute type; and for each attribute type, taking the maximum or minimum value among the multiple target sample values as the target threshold of the target object under the corresponding attribute type.
[0016] In one embodiment, a target object in the original scene is associated with multiple configuration components; the configuration components store the attribute value range and the attribute type corresponding to the attribute value range; the attribute value range includes an original threshold; updating the original scene according to the target threshold to obtain the target scene includes: finding the configuration component to be updated that stores the attribute type in the original scene; updating the original threshold of the attribute value range in the configuration component to be updated to the target threshold to obtain the updated configuration component; associating the updated configuration component with the corresponding target object to obtain the target scene.
[0017] In one embodiment, the method further includes: obtaining a scene sub-file corresponding to each attribute type in each sampling when performing attribute value sampling processing on each attribute value range; combining multiple scene sub-files to obtain a candidate scene file corresponding to each sampling value group; determining a target threshold corresponding to the target object from the candidate scene file, and updating the original scene file through the target threshold to obtain a target scene file; and storing the target scene file in a scene database.
[0018] In one embodiment, before obtaining the original scene file, the method further includes: obtaining preset configuration information, map information, and working condition information; the configuration information includes object identifiers and multiple configuration components; the working condition information includes obstacles and pedestrian interference; determining the target object corresponding to the object identifier, and constructing a test map for virtual testing based on the map information and the working condition information; determining the selected target configuration component from the multiple configuration components, and associating the target configuration component with the target object; the target configuration component includes multiple initial attribute types, and attribute information corresponding to each initial attribute type; configuring the test map according to the target object and the target configuration component to obtain the original scene corresponding to the test map.
[0019] Secondly, this application also provides a threshold-based scene generation apparatus. The apparatus includes:
[0020] The file acquisition module is used to acquire the original scene file containing the original scene of the target object; the original scene file includes multiple attribute types of the target object, and the attribute value range corresponding to each attribute type;
[0021] The set combination module is used to perform attribute value sampling processing on each of the attribute value intervals to obtain a sampling set corresponding to each attribute type; and to combine the sampled values in different sampling sets to obtain multiple sampled value groups;
[0022] The threshold determination module is used to filter out a target sample value group from the multiple sample value groups according to preset attribute value constraints; determine the target threshold of the target object under the corresponding attribute type according to the target sample value group; and update the original scene according to the target threshold to obtain the target scene.
[0023] Thirdly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, performs the following steps:
[0024] Obtain the original scene file containing the target object; the original scene file includes multiple attribute types of the target object, and the attribute value range corresponding to each attribute type;
[0025] Each attribute value range is sampled to obtain a sample set corresponding to each attribute type.
[0026] By combining the sampled values from different sample sets, multiple sampled value groups are obtained;
[0027] The target sample value group is selected from the multiple sample value groups according to the preset attribute value constraints;
[0028] Based on the target sample value group, the target threshold of the target object under the corresponding attribute type is determined, and the original scene is updated according to the target threshold to obtain the target scene.
[0029] The aforementioned threshold-based scene generation method, apparatus, and storage medium, by acquiring an original scene file containing the target object, can determine multiple attribute types of the target object and the corresponding attribute value ranges for each attribute type. Then, by sampling each attribute value range separately, a sampling set corresponding to each attribute type is obtained. By combining the sampled values from different sampling sets, multiple sample value groups are obtained. A target sample value group is then selected from these groups based on preset attribute value constraints. Thus, the target threshold for the target object under the corresponding attribute type can be determined based on the target sample value group. Updating the original scene according to the target threshold yields the target scene. Since this application performs attribute value sampling processing on each attribute value range before filtering different sample value groups, compared to the traditional method of manually editing the target object's attributes, this application can accurately obtain the target threshold for the target object under the corresponding attribute type. Furthermore, the attribute value sampling process can be viewed as an iterative process of an autonomous driving algorithm, thus achieving efficient modification and adaptation of the original scene file and improving the efficiency of target scene generation.
[0030] Furthermore, this application generates large-scale target scenes based on an original scene built from scratch that includes the running parameters of each target object. This means that multiple target scenes with similarities to the original scene can be derived, thus making the generated target scenes more diverse and meeting the needs of target scene diversity. Attached Figure Description
[0031] Figure 1 This is an application environment diagram of a threshold-based scene generation method in one embodiment;
[0032] Figure 2 This is a flowchart illustrating a threshold-based scene generation method in one embodiment;
[0033] Figure 3 This is a schematic diagram illustrating the principle of determining the target threshold in one embodiment;
[0034] Figure 4 This is a flowchart illustrating the process of determining the original scene in one embodiment;
[0035] Figure 5 This is a schematic diagram of a scene editing interface in one embodiment;
[0036] Figure 6 This is a structural block diagram of a threshold-based scene generation device in one embodiment;
[0037] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0038] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0039] The threshold-based scene generation method provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. Terminal 102 and server 104 can independently execute the threshold-based scene generation method provided in this application; they can also collaboratively execute the threshold-based scene generation method provided in this application. The following explanation uses the collaborative execution of the threshold-based scene generation method provided in this application by terminal 102 and server 104 as an example. Terminal 102 can send the acquired original scene file containing the target object to server 104. Server 104 is used to perform attribute value sampling processing on each attribute value range to obtain a sampling set corresponding to each attribute type, and combine the sampled values in different sampling sets to obtain multiple sampling value groups; according to preset attribute value constraints, a target sampling value group is selected from the multiple sampling value groups, and based on the target sampling value group, the target threshold of the target object under the corresponding attribute type is determined; server 104 is also used to update the original scene according to the target threshold to obtain the target scene, and send the target scene to terminal 102 so that terminal 102 can display the target scene. The terminal 102 may be, but is not limited to, various personal computers, laptops, smartphones, smart in-vehicle devices, tablets, IoT devices, and portable wearable devices. In one embodiment, the terminal 102 may be an autonomous driving computing platform that is not part of an autonomous driving vehicle. The server 104 may be implemented using a standalone server or a server cluster consisting of multiple servers.
[0040] In one embodiment, such as Figure 2 As shown, a threshold-based scene generation method is provided. The method is illustrated using a computer device as an example, which can provide... Figure 1 The terminal or server in the process includes the following steps:
[0041] Step 202: Obtain the original scene file containing the original scene of the target object.
[0042] In road testing based on the original scenario, the autonomous driving simulation system typically generates several target objects with interactive behaviors according to the configuration file or actual test requirements. The target objects include the main vehicle and the simulated objects. The main vehicle is an actual vehicle or virtual vehicle controlled by the autonomous driving algorithm. The simulated objects, also known as intelligent agents, are objects in the original scenario that may interact with the main vehicle, such as pedestrians, motor vehicles, bicycles, etc.
[0043] The original scene file can be a description file of the initial state and motion information of all target objects in an original scene; the original scene file includes multiple attribute types of target objects, and the attribute value range corresponding to each attribute type; the attribute type is a parameterized expression of a certain characteristic of the target object, and the attribute type can also be called an attribute item, such as the vehicle's driving speed, acceleration, the relative distance between the target object and the obstacle object, relative speed and other dynamic types; the attribute value range is the parameter range of a certain characteristic of the target object, which can usually be obtained from the configuration file edited by the user.
[0044] Specifically, the computer device can display a configuration interface generated by an editor, responding to the user's editing operations in the configuration interface to obtain the original scene file of the original scene, that is, to transform the intuitive operation of the original scene into a description file.
[0045] In one embodiment, the original scene file can be obtained through a scene descriptor, that is, by editing code scripts, the patterns of appearance, operation methods, and interactions with the simulated objects around the main vehicle can be defined.
[0046] Step 204: Perform attribute value sampling processing on each attribute value range to obtain the sampling set corresponding to each attribute type.
[0047] The attribute value range includes the sampling interval, the original minimum value, and the original maximum value; the original minimum value and the original maximum value are a kind of original threshold, which is usually specified by the user during the editing process.
[0048] Specifically, the computer device performs attribute value sampling on the original threshold through sampling intervals to obtain the sampling set corresponding to the attribute type.
[0049] In one embodiment, attribute value sampling is performed on each attribute value range to obtain a sampling set corresponding to each attribute type. This includes: for each attribute type among multiple attribute types, starting from the original minimum value, attribute value sampling is performed according to the sampling interval until the original maximum value is sampled to obtain the sampling value corresponding to each sampling; and multiple sampling values are combined to obtain a sampling set corresponding to each attribute type.
[0050] Specifically, the range of attribute values usually differs depending on the attribute type. An attribute value range can be considered a special type of field, such as... Figure 3 As shown, Figure 3 This is a schematic diagram illustrating the principle of determining the target threshold. The computer device iterates through the fields of a specific item and samples attribute values according to sampling intervals, thus obtaining multiple sampled values, including the original minimum and maximum values.
[0051] For example, in a scenario where the target object is a main vehicle that needs to change lanes to the left, and the obstacle in the left lane is a simulated object moving forward, it is necessary to determine the range of relative distances between the target object and the obstacle object at different speeds for the lane change to succeed. (Reference) Figure 3 For example, let's consider speed as Data1, with a sampling interval of 0.5, an initial minimum value of 5, and an initial maximum value of 6; and relative distance as Data2, with a sampling interval of 4, an initial minimum value of 8, and an initial maximum value of 16. Therefore, by sampling each attribute value interval, we obtain sampled values of 5, 5.5, and 6 in sample set 1 corresponding to speed, and 8, 12, and 16 in sample set 2 corresponding to relative distance. It is easy to understand that the data size of the sampled values and the attribute types corresponding to the sampled values described in this application are merely examples and do not have the same meaning as actual test scenarios.
[0052] Step 206: Combine the sampled values from different sample sets to obtain multiple sampled value groups.
[0053] Specifically, for each of the multiple sample sets, the computer device combines each sample value in the current sample set with each sample value in the other sample sets (excluding the current sample set) to obtain multiple groups of sample values. For example, refer to... Figure 3 The sample value 5 in sample set 1 and the sample value 8 in sample set 2 are combined to obtain sample value group 1; the sample value 5.5 in sample set 1 and the sample value 8 in sample set 2 are combined to obtain sample value group 2.
[0054] In one embodiment, sampled values from different sample sets are combined to obtain multiple sampled value groups, including: combining sampled value M from sample set i. (i，n) With sample value M in sample set j (j，k) By combining them, multiple sample value groups are obtained; i∈(1,2…N), j∈(1,2…N), N represents any one of the multiple sample sets; n, k, and N are all positive integers.
[0055] Step 208: Select the target sample value group from multiple sample value groups according to the preset attribute value constraints.
[0056] Among them, attribute value constraints are used to filter sampled values in the sampled value group based on factors such as interaction logic and traffic rules in the virtual scene. Attribute value constraints are obtained through the autonomous driving evaluation system, and users can edit different attribute value constraints in the autonomous driving evaluation system. For example, the maximum speed on a highway cannot exceed the prescribed speed, and when a target object encounters a red light while merging into an obstacle object, both the target object and the obstacle object must slow down.
[0057] Specifically, the computer equipment obtains attribute value constraints from the autonomous driving evaluation system and determines whether each sample value in multiple sample value groups meets the attribute value constraints. If any sample value in a sample value group does not meet the attribute value constraints, then that sample value group is considered a filterable sample value group; if each sample value in a sample value group meets the attribute value constraints, then that sample value group is considered the target sample value group. For example, refer to... Figure 3 Sample value group 1, sample value group 2, sample value group 4 and sample value group 5 are selected as target sample value groups.
[0058] In one embodiment, if the target sample value group cannot be selected from multiple sample value groups, it can be considered that the attribute value range corresponding to the attribute type is set too strictly, and the user needs to re-edit the configuration file.
[0059] Step 210: Based on the target sample value group, determine the target threshold of the target object under the corresponding attribute type, and update the original scene according to the target threshold to obtain the target scene.
[0060] The target sample value group selected from multiple sample value groups usually includes multiple groups.
[0061] In one embodiment, determining the target threshold of the target object under the corresponding attribute type based on the target sample value group includes: clustering the sample values in the target sample value group according to the attribute type to obtain multiple target sample values corresponding to each attribute type; and for each attribute type, taking the maximum or minimum value among the multiple target sample values as the target threshold of the target object under the corresponding attribute type.
[0062] Specifically, the computer device clusters the sampled values in multiple target sampled value groups according to attribute type, obtaining multiple target sampled values corresponding to each attribute type. For example, for the attribute type "speed Data1", the sampled values in sampled value groups 1, 2, 4, and 5 are clustered to obtain the target sampled values 5, 5.5, 5, and 5.5 corresponding to speed Data1. Based on the interaction logic and traffic rules in the virtual scene, the computer device can use the maximum or minimum value among the multiple target sampled values as the target threshold for the corresponding attribute type. For example, the maximum value of 5.5 among the target sampled values corresponding to speed Data1 can be used as the target threshold, meaning the target object's driving speed can be less than 5.5. A driving speed of 0 can be considered as the target object slowing down until it stops to avoid danger. Alternatively, both the minimum value of 5 and the maximum value of 5.5 can be used as the target threshold.
[0063] In one embodiment, when the target sample value group selected from multiple sample value groups is only one, the sample value in the target sample value group is directly used as the target threshold of the target object under the corresponding attribute type.
[0064] In one embodiment, the computer device can determine at least one candidate threshold for each attribute type based on multiple target sample values and a target threshold corresponding to each attribute type. For example, the candidate thresholds for speed Data1 could be 5.4, 5.2, etc.
[0065] In one embodiment, updating the original scene according to a target threshold to obtain a target scene includes: finding the configuration component to be updated that stores attribute types in the original scene; updating the original threshold of the attribute value range in the configuration component to be updated to the target threshold to obtain an updated configuration component; and associating the updated configuration component with the corresponding target object to obtain the target scene.
[0066] In the original scene, the target object is associated with multiple configuration components; the configuration components store attribute value ranges and the attribute types corresponding to the attribute value ranges; the attribute value ranges include the original threshold.
[0067] Specifically, the original scene is obtained by associating and configuring the target object with multiple configuration components. These configuration components indicate the behavioral characteristics of the target object within the original scene. Once a configuration component is associated with the target object, the target object can be configured using the information carried within it. The attribute types and value ranges in the original scene file can be represented by different configuration components; thus, the attribute types and value ranges can be considered as information stored within the configuration components. Therefore, for each target threshold, the computer device searches the original scene for the configuration component storing the attribute type corresponding to the target threshold, updates the original threshold of the attribute value range in the configuration component to the target threshold, and obtains the updated configuration component. The computer device then associates the updated configuration component with the corresponding target object, thereby obtaining the target scene based on the original scene.
[0068] In one embodiment, the scene category of the target scene is the same as that of the original scene. The scene category refers to the scene type or event type. The scene type is related to traffic rules, such as following another vehicle, changing lanes, or intersections. The event type may include rear-end collisions, sudden braking, speeding, etc.
[0069] In one embodiment, the computer device can update the original thresholds in the configuration components to be updated to different candidate thresholds, thereby obtaining multiple updated configuration components corresponding to the target object. It can then select any one of these updated configuration components and associate it with the target object. This allows for a richer variety of generated target scenes, meeting the needs of diverse target scene requirements.
[0070] The aforementioned threshold-based scene generation method obtains an original scene file containing the target object, determines multiple attribute types of the target object, and the corresponding attribute value ranges for each attribute type. Then, by sampling each attribute value range, a sampling set corresponding to each attribute type is obtained. By combining the sampled values from different sampling sets, multiple sample value groups are obtained. A target sample value group is selected from these groups based on preset attribute value constraints. Thus, the target threshold for the target object under the corresponding attribute type can be determined based on the target sample value group. Updating the original scene according to the target threshold yields the target scene. Since this application samples each attribute value range before filtering different sample value groups, compared to the traditional method of manually editing the target object's attributes, this application can accurately obtain the target threshold for the target object under the corresponding attribute type. Furthermore, the attribute value sampling process can be viewed as an iterative process of an autonomous driving algorithm, thus achieving efficient modification and adaptation of the original scene file and improving the efficiency of target scene generation.
[0071] In one embodiment, before obtaining the original scene file containing the target object, the method further includes: obtaining the original scene file; traversing the fields of the attribute information corresponding to each initial attribute type; if the attribute information currently traversed includes attribute constants, then determining that the initial attribute type is an attribute type that does not require threshold adjustment; if the attribute information currently traversed includes attribute value ranges, then determining that the initial attribute type is an attribute type that requires threshold adjustment.
[0072] The original scene file includes multiple initial attribute types and corresponding attribute information for each initial attribute type. The initial attribute types include dynamic types such as vehicle speed, as well as preset constant types such as vehicle length, width, height, and initial position. Attribute information includes at least one of attribute value ranges and attribute constants. Attribute value ranges correspond to the initial attribute types of dynamic types, and attribute constants correspond to the initial attribute types of constant types. Attribute information can be presented in a field format.
[0073] Specifically, the computer device iterates through the fields of each attribute information, determining whether each attribute information includes an attribute value range. If it does, the attribute type is determined to be dynamic, meaning the original threshold in the attribute value range needs to be adjusted. If it doesn't, the attribute type is determined to be constant, and the attribute information corresponding to this attribute type is an attribute constant. For example, refer to... Figure 3 For example, if the vehicle length is Data3, the attribute constant corresponding to the vehicle length is 120.
[0074] In one embodiment, when the attribute information corresponding to the initial attribute type is an attribute constant, the original scene does not need to be generated through algorithm iteration.
[0075] In one embodiment, the initial attribute types and attribute information can be presented through different configuration components, which are used to indicate the behavioral characteristics of the target object in the original scene. Once a configuration component is associated with the target object, the target object can be configured using the initial attribute types and attribute information carried in the configuration component.
[0076] In this embodiment, by traversing the attribute information in advance, the initial attribute type belonging to the dynamic type can be determined, thereby ensuring that the attribute value range can be accurately sampled and processed in the future. This avoids the problem of wasting resources when the algorithm is still iterated on the original scene because the initial attribute type corresponds to the attribute constant.
[0077] In one embodiment, selecting a target sample value group from multiple sample value groups based on preset attribute value constraints includes: for each sample value group in the multiple sample value groups, determining whether the behavioral characteristics of the target object conform to traffic rules by using each sample value in the current sample value group corresponding to the target object; when the behavioral characteristics conform to traffic rules, determining whether the behavioral characteristics of the target object will experience a sudden situation; when no sudden situation occurs, using the current sample value group as the target sample value group.
[0078] Among them, the attribute value constraints include at least traffic rules and emergencies; emergencies include at least one of collision and rear-end collision.
[0079] Specifically, since the initial attribute types and attribute information can be presented through different configuration components, these components indicate the behavioral characteristics of the target object in the original scene. That is, the multiple attribute types corresponding to the target object, and the sampled values corresponding to each attribute type, all reflect the behavioral characteristics of the target object. Therefore, the computer device can determine the behavioral characteristics of the target object based on each sampled value in the sampled value group, and determine whether the behavioral characteristics comply with traffic rules. For example, refer to... Figure 3 The speed in sampling value group 3 corresponds to a sampling value of 6. Given that the speed limit on highways is 5.8, the behavioral characteristics reflected in sampling value group 3 do not comply with the traffic rules. The speed in sampling value group 7 corresponds to a sampling value of 5.8. This complies with the speed limit on highways, but the target object was found to have run a red light. Therefore, neither sampling value group 3 nor sampling value group 7 can be used as the target sampling value group.
[0080] Furthermore, when the target object's behavioral characteristics are determined to conform to traffic rules, the computer equipment continues to determine whether any unforeseen circumstances will occur. Only when no unforeseen circumstances occur, i.e., all attribute value constraints are met, is the current sampled value group used as the target sampled value group. For example, refer to... Figure 3 The speed sample value in sample value group 8 is 5.5, which complies with the traffic rule that the speed limit on the highway is 5.8. There is no behavior characteristic of running a red light. However, a collision between the target object and the obstacle object was detected. Therefore, sample value group 8 cannot be used as the target sample value group.
[0081] In this embodiment, the sampled value groups corresponding to multiple attribute types are filtered by attribute value constraints, so that each sampled value in the target sampled value group conforms to the interaction rules in the virtual scene, providing a selection basis for further determining the target threshold of the target object under the corresponding attribute type.
[0082] In one embodiment, the method further includes: obtaining a scene sub-file corresponding to each attribute type at each sampling time when performing attribute value sampling processing on each attribute value range; combining multiple scene sub-files to obtain a candidate scene file corresponding to each sampling value group; determining the target threshold corresponding to the target object from the candidate scene file, and updating the original scene file through the target threshold to obtain the target scene file; and storing the target scene file in the scene database.
[0083] Among them, the candidate scene file corresponds to the sample set corresponding to the same attribute type, and the candidate scene file corresponds to the sample value group obtained by combining different sample sets.
[0084] Specifically, the computer device may include a scene generator. By using the original scene file as the input template and sampling each attribute value range in the original scene file—that is, an iterative process using an autonomous driving algorithm—a scene sub-file corresponding to each attribute type at each sampling can be obtained. The specific implementation process can be found in step 204. Since each attribute value sampling yields a scene sub-file for each attribute type at the current sampling time, the computer device can combine multiple scene sub-files to obtain candidate scene files corresponding to each sampling value group. The specific implementation process can be found in step 206. Since the candidate scene file corresponding to the current sampling can be considered a separate scene file, it can be used as the algorithm iteration basis for obtaining the candidate scene file corresponding to the next sampling. The computer device determines the target threshold corresponding to the target object from the candidate scene file and updates the original scene file using the target threshold to obtain the target scene file. The specific implementation process can be found in step 210, which will not be elaborated further in this application.
[0085] In this embodiment, by treating the attribute value sampling process as an iterative process of an autonomous driving algorithm, the candidate scene files at different sampling times can be automatically generated as the algorithm iterates without the need for manual modification. Therefore, efficient modification and adaptation of the original scene files are achieved, improving the efficiency of target scene file generation.
[0086] In one embodiment, such as Figure 4 As shown, the process of generating the original scene before obtaining the original scene file includes the following steps:
[0087] Step 402: Obtain preset configuration information, map information, and operating condition information.
[0088] The configuration information includes object identifiers and multiple configuration components. These components predefine different behavioral characteristics that can be associated with the target object, allowing configuration of the target object through these behavioral characteristics. These behavioral characteristics can be presented through multiple initial attribute types and their corresponding attribute information. Map information may include road labeling information, such as road areas and lane line types. Map information can be obtained through user-preset road arrangement rules or edited using an editor system. Operating condition information may include obstacles, traffic lights, and road signs.
[0089] Step 404: Determine the target object corresponding to the object identifier, and construct a test map for virtual testing based on map information and working condition information.
[0090] The object identifier can be an ID number, used to uniquely identify the target object.
[0091] Specifically, such as Figure 5 As shown, Figure 5 This is a schematic diagram of a scene editing interface. The computer device constructs a test map for virtual testing based on map information and operational data, and in response to user editing operations within the scene editing interface. For example, road signs can be added within the scene editing interface.
[0092] Step 406: Determine the selected target configuration component from multiple configuration components and associate the target configuration component with the target object.
[0093] The target configuration component includes multiple initial attribute types and attribute information corresponding to each initial attribute type. In other words, the target configuration component can be used to indicate the behavioral characteristics of the target object in the original scene.
[0094] Specifically, the computer device can display an information configuration interface including multiple configuration components. When the computer device responds to a user's selection operation on these components, it determines the target configuration component selected. These configuration components are stored in a predefined component library. For model components within the target configuration component, the computer device can import them from a third-party model library. When the computer device associates the target configuration component with a target object, the target object can be configured with all the behavioral characteristics of the target configuration component.
[0095] In one embodiment, before associating a target object with a target configuration component, the computer device may treat the target object as an empty object, that is, without any features. The interactive behavior of the target object in the target scene to be generated will also be different depending on the components mounted in the empty object.
[0096] Step 408: Configure the test map according to the target object and target configuration components to obtain the original scene corresponding to the test map.
[0097] The target configuration component includes at least one of a transformation component, a behavior component, and a collision component. The transformation component represents the initial position of the target object on the test map. The behavior component may include a straight-moving component and an avoidance component. The straight-moving component represents that the target object will move forward in a straight line when there are no factors such as traffic lights. The avoidance component represents that when encountering obstacles or other factors, the target object will avoid them based on specific historical information, such as deceleration avoidance or lane change avoidance.
[0098] Specifically, the computer device determines the initial position of the target object on the test map based on the transformation component and places the target object at the initial position. Then, based on the behavior component, it determines the target behavior of the target object after it starts from the initial position. The target behavior includes at least one of two actions: a straight-line behavior and an avoidance behavior. The straight-line behavior corresponds to the straight-line component, and the avoidance behavior corresponds to the avoidance component. The computer device configures the target object through a collision component to obtain the original scene. The collision component characterizes whether the target object is a rigid body and whether it is transformed into a suitable physical model upon collision. The collision features corresponding to the collision component can include simple bending and severe deformation, etc.
[0099] In one embodiment, the behavioral characteristics in the target configuration component may further include initial speed and direction of travel, thereby ensuring that the target object begins to travel in the original scene according to the initial speed and direction of travel.
[0100] In this embodiment, by associating the target configuration component with the target object, the behavioral characteristics of the target object in the original scene can be determined. Therefore, when different target configuration components are used to modify the behavioral characteristics of the target object, the interactive behavior of different target objects can be flexibly defined, thereby improving the generation efficiency of the original scene.
[0101] In one embodiment, the computer device acquires an original scene file of the original scene, wherein the original scene file includes multiple attribute types of the target object and an attribute value range corresponding to each attribute type. The attribute value range includes a sampling interval, an original minimum value, and an original maximum value. For each attribute type, the computer device samples attribute values starting from the original minimum value and according to the sampling interval until the original maximum value is sampled, obtaining the sampled value corresponding to each sampling. The computer device then combines the sampled values from different sampled sets to obtain multiple sampled value groups, and performs sampling on each sampled value in the multiple sampled value groups. Each group uses each sample value in the current sample value group corresponding to the target object to determine whether the target object's behavior characteristics conform to preset traffic rules. When the behavior characteristics conform to the preset traffic rules, the computer device determines whether the target object's behavior characteristics will cause a sudden situation. If no sudden situation occurs, the current sample value group is used as the target sample value group. The computer device clusters the sample values in the target sample value group according to the attribute type to obtain multiple target sample values corresponding to each attribute type. For each attribute type, the maximum or minimum value among the multiple target sample values is used as the target threshold of the target object under the corresponding attribute type. This allows the computer device to update the original scene according to the target threshold to obtain the target scene.
[0102] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0103] Based on the same inventive concept, this application also provides a threshold-based scene generation apparatus for implementing the threshold-based scene generation method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more threshold-based scene generation apparatus embodiments provided below can be found in the limitations of the threshold-based scene generation method described above, and will not be repeated here.
[0104] In one embodiment, such as Figure 6As shown, a threshold-based scene generation device 600 is provided, including: a file acquisition module 602, a set combination module 604, and a threshold determination module 606, wherein:
[0105] The file acquisition module 602 is used to acquire the original scene file containing the original scene of the target object; the original scene file includes multiple attribute types of the target object, and the attribute value range corresponding to each attribute type.
[0106] The set combination module 604 is used to perform attribute value sampling processing on each attribute value range to obtain the sampling set corresponding to each attribute type; and to combine the sampled values in different sampling sets to obtain multiple sampled value groups.
[0107] The threshold determination module 606 is used to filter out the target sample value group from multiple sample value groups according to the preset attribute value constraints; determine the target threshold of the target object under the corresponding attribute type according to the target sample value group; and update the original scene according to the target threshold to obtain the target scene.
[0108] In one embodiment, the threshold-based scene generation device 600 further includes a judgment module 608, used to obtain an original scene file; the original scene file includes multiple initial attribute types and attribute information corresponding to each initial attribute type; the attribute information includes at least one of attribute value range and attribute constant; the field traversal is performed on the attribute information corresponding to each initial attribute type; if the attribute information currently traversed includes attribute constant, then the initial attribute type is determined to be an attribute type that does not require threshold adjustment; if the attribute information currently traversed includes attribute value range, then the initial attribute type is determined to be an attribute type that requires threshold adjustment.
[0109] In one embodiment, the set combination module 604 is further configured to, for each of the multiple attribute types, start from the original minimum value and sample the attribute value according to the sampling interval until the original maximum value is sampled, to obtain the sampled value corresponding to each sampling; and combine the multiple sampled values to obtain the sample set corresponding to each attribute type.
[0110] In one embodiment, the set combination module 604 is further configured to combine the nth sample value M from the i-th sample set. (i，n) With sample value M in sample set j (j，k) By combining them, multiple sample value groups are obtained; i∈(1,2…N), j∈(1,2…N), N represents any one of the multiple sample sets; n, k, and N are all positive integers.
[0111] In one embodiment, the threshold determination module 606 includes a filtering module 6061, which is used to determine whether the behavioral characteristics of the target object conform to preset traffic rules for each of the multiple sampling value groups by using each sampling value in the current sampling value group corresponding to the target object; when the behavioral characteristics conform to the preset traffic rules, it is determined whether the behavioral characteristics of the target object will cause a sudden situation; when no sudden situation occurs, the current sampling value group is used as the target sampling value group.
[0112] In one embodiment, the threshold determination module 606 further includes a clustering module 6062, which is further configured to cluster the sampled values in the target sampled value group according to the attribute type to obtain multiple target sampled values corresponding to each attribute type; for each attribute type, the maximum or minimum value among the multiple target sampled values is used as the target threshold of the target object under the corresponding attribute type.
[0113] In one embodiment, the threshold determination module 606 further includes an update module 6063, which is further configured to: search for the configuration component to be updated that stores the attribute type in the original scene; update the original threshold of the attribute value range in the configuration component to be updated to the target threshold to obtain the updated configuration component; and associate the updated configuration component with the corresponding target object to obtain the target scene.
[0114] In one embodiment, the threshold-based scene generation device 600 further includes a file update module 610, which is used to obtain a scene sub-file corresponding to each attribute type at each sampling time when performing attribute value sampling processing on each attribute value range; combine multiple scene sub-files to obtain a candidate scene file corresponding to each sampling value group; determine the target threshold corresponding to the target object from the candidate scene file, and update the original scene file with the target threshold to obtain the target scene file; and store the target scene file in the scene database.
[0115] In one embodiment, the threshold-based scene generation device 600 further includes an original scene determination module 612, used to acquire preset configuration information, map information, and working condition information; the configuration information includes an object identifier and multiple configuration components; determine the target object corresponding to the object identifier, and construct a test map for virtual testing based on the map information and working condition information; determine the selected target configuration component from the multiple configuration components, and associate the target configuration component with the target object; the target configuration component includes multiple initial attribute types, and attribute information corresponding to each initial attribute type; perform initial configuration on the test map based on the target object and the target configuration component to obtain the original scene corresponding to the test map.
[0116] Each module in the threshold-based scene generation device described above can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0117] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, input / output interfaces (I / O), a communication interface, a display unit, and input devices. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface, display unit, and input devices are also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a threshold-based scene generation method. The display unit of the computer device is used to form a visually visible image. It can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.
[0118] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0119] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0120] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0121] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0122] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A threshold-based scene generation method, characterized in that, The method includes: Obtain an original scene file containing the original scene of the target object; the original scene file includes multiple attribute types of the target object and attribute value ranges corresponding to each attribute type; the target object is associated with multiple configuration components, and the configuration components store the attribute value ranges and the attribute types corresponding to the attribute value ranges; the target object is at least one of a main vehicle and a simulated object; the attribute type is at least one of the vehicle's driving speed, acceleration, the relative distance between the target object and the obstacle object, and the relative speed between the target object and the obstacle object; Each attribute value range is sampled to obtain a sample set corresponding to each attribute type. By combining the sampled values from different sample sets, multiple sampled value groups are obtained; Multiple target sample value groups are selected from the multiple sample value groups according to preset attribute value constraints; Cluster the sampled values in the plurality of target sampled value groups according to the attribute type to obtain a plurality of target sampled values corresponding to each attribute type. For each attribute type, the maximum or minimum value among the plurality of target sampled values is used as the target threshold of the target object under the corresponding attribute type. The original scene is searched for the configuration component to be updated that stores the attribute type. The original threshold of the attribute value range in the configuration component to be updated is updated to the target threshold to obtain the updated configuration component. The updated configuration component is associated with the corresponding target object to obtain the target scene.
2. The method according to claim 1, characterized in that, Before obtaining the original scene file containing the original scene of the target object, the method further includes: Obtain the original scene file; the original scene file includes multiple initial attribute types and attribute information corresponding to each initial attribute type; the attribute information includes at least one of the attribute value range and attribute constant; For each of the initial attribute types, perform field traversal on the corresponding attribute information; If the attribute information currently being traversed includes the attribute constant, then the initial attribute type is determined to be an attribute type that does not require threshold adjustment. If the attribute information currently being traversed includes the attribute value range, then the initial attribute type is determined to be an attribute type that requires threshold adjustment.
3. The method according to claim 1, characterized in that, The attribute value range includes a sampling interval, an original minimum value, and an original maximum value; the step of performing attribute value sampling processing on each of the attribute value ranges to obtain a sampling set corresponding to each attribute type includes: For each of the multiple attribute types, starting from the original minimum value, attribute values are sampled according to the sampling interval until the original maximum value is sampled, thus obtaining the sampled value corresponding to each sampling. By combining multiple sample values, a sample set corresponding to each attribute type is obtained.
4. The method according to claim 1, characterized in that, The step of combining sampled values from different sample sets to obtain multiple sampled value groups includes: Take the nth sample value M from the i-th sample set. （i，n） The sample value M in the sample set described by j is k. （j，k） By combining them, multiple sample value groups are obtained; i∈(1,2…N), j∈(1,2…N), N represents any one of the multiple sample sets; n, k, and N are all positive integers.
5. The method according to claim 1, characterized in that, The attribute value constraints are obtained through an autonomous driving evaluation system; the attribute value constraints include traffic rules and unexpected situations; the step of selecting multiple target sample value groups from the multiple sample value groups according to the preset attribute value constraints includes: For each of the multiple sample value groups, the behavior characteristics of the target object are determined to conform to traffic rules by using each sample value in the current sample value group corresponding to the target object; When the behavioral characteristics conform to traffic rules, it is determined whether the behavioral characteristics of the target object will lead to a sudden situation; the sudden situation includes at least one of collision and rear-end collision. If the aforementioned emergency does not occur, the current sampled value group will be used as the target sampled value group.
6. The method according to claim 1, characterized in that, The method further includes: When performing attribute value sampling processing on each of the attribute value ranges, the scene sub-file corresponding to each attribute type is obtained for each sampling. Multiple scene sub-files are combined to obtain candidate scene files corresponding to each sample value group; The target threshold corresponding to the target object is determined from the candidate scene file, and the original scene file is updated using the target threshold to obtain the target scene file; The target scene file is stored in the scene database.
7. The method according to claim 2, characterized in that, Before obtaining the original scene file, the method further includes: Obtain preset configuration information, map information, and operating condition information; the configuration information includes object identifiers and multiple configuration components; the operating condition information includes at least one of obstacles and traffic lights; The target object corresponding to the object identifier is determined, and a test map for virtual testing is constructed based on the map information and the working condition information; The selected target configuration component is determined from multiple configuration components, and the target configuration component is associated with the target object; the target configuration component includes multiple initial attribute types and attribute information corresponding to each initial attribute type; Based on the target object and the target configuration component, the test map is configured to obtain the original scene corresponding to the test map.
8. A threshold-based scene generation device, characterized in that, The device includes: The file acquisition module is used to acquire an original scene file containing the original scene of the target object; the original scene file includes multiple attribute types of the target object, and the attribute value range corresponding to each attribute type; the target object is at least one of a main vehicle and a simulated object; the attribute type is at least one of the vehicle's driving speed, acceleration, the relative distance between the target object and the obstacle object, and the relative speed between the target object and the obstacle object; The set combination module is used to sample the attribute values for each of the attribute value ranges to obtain a sample set corresponding to each attribute type; and to combine the sampled values in different sample sets to obtain multiple sample value groups. The threshold determination module is used to filter out multiple target sample value groups from the multiple sample value groups according to preset attribute value constraints; cluster the sample values in the multiple target sample value groups according to the attribute type to obtain multiple target sample values corresponding to each attribute type; for each attribute type, the maximum or minimum value among the multiple target sample values is used as the target threshold of the target object under the corresponding attribute type; search for the configuration component to be updated that stores the attribute type in the original scene; update the original threshold of the attribute value range in the configuration component to be updated to the target threshold to obtain the updated configuration component; associate the updated configuration component with the corresponding target object to obtain the target scene.
9. The apparatus according to claim 8, characterized in that, The device further includes: A judgment module is used to obtain an original scene file; the original scene file includes multiple initial attribute types and attribute information corresponding to each initial attribute type; the attribute information includes at least one of the attribute value range and attribute constant; the module iterates through the attribute information corresponding to each initial attribute type; if the attribute constant is included in the currently traversed attribute information, the initial attribute type is determined to be an attribute type that does not require threshold adjustment; if the attribute value range is included in the currently traversed attribute information, the initial attribute type is determined to be an attribute type that requires threshold adjustment.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.